Employee turnover refers to the replacement of employees within an organization, which can lead to losses such as recruitment costs and decreased productivity. Predicting turnover is crucial for companies to anticipate and take appropriate actions to retain potential employees. This study aims to optimize the employee turnover prediction model by integrating hash encoding techniques and machine learning. The dataset used in this study is an open-source dataset obtained from Kaggle dataset. It consists of 14,994 rows and 10 columns (features) representing employee-related information such as satisfaction level, evaluation score, number of projects, average monthly hours, and whether the employee left the company. Among these features, some are of object data type. Since machine learning algorithms generally cannot work directly with object-type features, the use of hash encoding is proposed. This technique converts object-type data into numerical data. It is part of the preprocessing stage, aiming to reduce memory usage, speed up data preprocessing, and improve model performance. After preprocessing is completed, the prediction model is trained using the Random Forest algorithm to predict employee turnover. The evaluation is conducted using accuracy, recall, precision, and F1-score metrics, which yielded results of 0.988, 0.961, 0.988, and 0.974, respectively. These results indicate that the integration of hash encoding techniques and machine learning can produce a well-performing model for predicting employee turnover.

This study examines the performance of Random Forest and XGBoost in predicting the diagnosis and severity of respiratory diseases using a simulated dataset of 2,000 patient records. The models were tested on two classification tasks: identifying disease types (e.g., pneumonia, influenza) and classifying severity levels (mild, moderate, severe). Both models achieved perfect accuracy in severity classification, with 1.0000 ± 0.0000 cross-validation scores, demonstrating strong stability under balanced class distributions. However, in the diagnosis task, Random Forest underperformed on minority classes, particularly pneumonia, with a recall of 0.18 and F1-score of 0.31. XGBoost, on the other hand, achieved superior results across all classes, including minority cases, with 0.9825 ± 0.0170 cross-validation accuracy and perfect test set performance. These findings highlight XGBoost’s robustness in handling imbalanced and multiclass medical data, making it a promising candidate for clinical decision support. Future work should address class imbalance and explore explainability techniques to improve trust and transparency in real-world applications.

Healthcare resource allocation remains a persistent challenge in Botswana, primarily due to inefficiencies in data management that obstruct equitable distribution and evidence-based decision-making. Traditional allocation approaches in Botswana exhibit severe fragmentation, low interoperability, and an absence of real-time data analytics factors that contribute to service delivery disparities, especially in rural and underserved areas. In contrast, developed countries have leveraged data warehousing to optimize healthcare resource planning, offering Botswana a proven yet untapped strategic opportunity. This study designs and validates a context-sensitive data warehouse methodology, applying the Kimball Lifecycle model as the guiding framework. A mixed-methods design was adopted, incorporating qualitative interviews with 24 healthcare practitioners and administrators across public and private health facilities, along with quantitative surveys assessing the state of 12 existing health data systems. Results reveal systemic shortcomings in data accuracy (average error rates of 22%), timeliness (with a median data update lag of 14 days), and accessibility (only 38% of facilities had centralized access). Post-implementation of the prototype data warehouse, significant improvements were noted: data accuracy increased by 47%, data accessibility across departments rose to 85%, and decision turnaround time was reduced by 33%. The warehousing also demonstrated cost-effectiveness, reducing redundant data handling expenses by an estimated 18% over six months. In conclusion, this study presents a robust, scalable, and locally adaptable data warehousing framework that effectively addresses Botswana’s systemic challenges in healthcare resource allocation.

Blockchain technology presents transformative opportunities for secure personal data sharing, particularly in healthcare, finance, and identity management. However, its widespread adoption is constrained by challenges such as limited scalability, privacy concerns, and conflicts with regulatory frameworks like the General Data Protection Regulation (GDPR). This study introduces a novel hybrid framework that integrates the InterPlanetary File System (IPFS) for off-chain storage with Zero-Knowledge Proofs (ZKPs) to enhance privacy, ensure regulatory compliance, and reduce on-chain storage demands. Employing a Design Science Research (DSR) methodology, the framework was developed and validated using Ethereum and Hyperledger Fabric, guided by insights from a systematic review of 180 studies from 2018 to 2023. Empirical evaluations revealed a 75% reduction in blockchain storage, 98% GDPR compliance, and zk-SNARK proof verification times below one second. The framework also enables GDPR-compliant erasure by removing encrypted off-chain data while preserving on-chain auditability. Despite challenges such as IPFS latency and trusted setup complexities, the solution offers a scalable and privacy-preserving architecture applicable to real-world domains, especially in privacy-critical environments like healthcare and finance by resolving blockchain’s GDPR compliance paradox.

This research applies sentiment analysis to understand user perceptions of the Access by KAI application, especially specific aspects such as speed, payment process, and user interface (UI/UX). User reviews are collected and processed through preprocessing stages, balancing using the SMOTE method, and classified using three machine learning algorithms, namely Support Vector Machine (SVM), Decision Tree, and Logistic Regression. The SVM model achieved the highest accuracy of 89.33%, followed by Logistic Regression at 88%, and Decision Tree at 86.67%. Precision, recall, and F1-scores for each model were also evaluated, showing strong performance in detecting negative sentiments but lower performance for neutral and positive sentiments. In addition, keyword-based analysis revealed that negative sentiment was most commonly found in the aspects of the payment process and speed. WordCloud visualization also strengthens the results by showing the dominance of negative words in user reviews. The results of this study provide important suggestions and input for application developers to improve aspects of the service that are considered less satisfactory by users. Thus, this study can be used as a practical guide in making strategic decisions to improve the quality of service and user satisfaction of the Access by KAI application.

Taxicab entrepreneurs who operate on e-hailing platforms in South Africa face challenges such as earning below minimum wage, lacking employment benefits, working long hours, and experiencing victimisation by traditional taxicab operators. The key question is why these entrepreneurs continue using e-hailing platforms despite unfavourable working conditions. This study proposed that technology adoption factors enable entrepreneurs to overcome challenges and encourage them to keep using e-hailing platforms. Based on this assumption, this study investigated the determinants of technology adoption that influence the attitude of taxicab entrepreneurs to continue using e-hailing platforms in South Africa. The researchers gathered quantitative data from 253 entrepreneurs in Johannesburg, South Africa and tested the hypotheses with multiple regression analysis. The results demonstrated that perceived usefulness, benefits, and security strongly influenced entrepreneurs' willingness to continue operating on e-hailing platforms. However, perceptions of convenience, trust, and perceived ease of use did not affect their decision to use e-hailing services. Theoretically, this study pinpointed the factors that drive and hinder the continued use of e-hailing applications. Practically, the results provide insights into understanding long-term usage, user satisfaction, and the success of e-hailing in developing countries undergoing digital transformation, such as South Africa.

This research explores the implementation of machine learning to predict depression among university students using a dataset of 2.028 responses containing PHQ-9 scores and academic-demographic attributes. The research implements a structured modeling process involving feature selection, normalization, the model’s efficacy was gauged through a suite of evaluate measures, encompassing accuracy, precision, recall, F1-score, The support vector machine (SVM) model’s accuracy improved from 58.8% to 99.5% after hyperparameter tuning. This investigation lends itself to the advancement of a proactive identification framework, which hold potential for incorporation within collegiate mental well-being surveillance infrastructures. Future implementations may consider real-time models and expand data sources through digital counseling systems and behavioral analytics

The growing volume of electronic waste (e-waste) in Tanzanian public institutions poses serious cybersecurity risks, as discarded devices often contain sensitive data vulnerable to unauthorized access. This study examines these risks across 11 public institutions, involving IT staff, e-waste handlers, policymakers, and environmental officers. It applies Routine Activity Theory, a framework that explains risks as arising when cybercriminals exploit unsecured e-waste due to weak regulations. Through interviews and focus group discussions, the research identifies key vulnerabilities: data leakage from improper sanitization, regulatory gaps, and risks from informal disposal methods like auctions. These findings highlight the need for stronger oversight to prevent data breaches. The study proposes a framework that categorizes devices by risk level and integrates secure sanitization protocols, such as data wiping or destruction. Policymakers and institutions must urgently adopt these protocols to protect sensitive data and promote sustainable e-waste management in Tanzania’s public sector.

Modern project management in organisations is moving towards Scaled Agile to achieve success. Scaled Agile refers to a set of organisational structures and processes for implementing agile practices that are applied on an enterprise scale. This study explores Scaled Agile growth and impact by analysing 238 publications obtained from the Scopus databases using bibliometric analysis. The results show that publications on Scaled Agile have steadily increased, with more contributions from developed nations than developing countries. In terms of the geographic distribution of publications, Germany is the leading followed by Sweden and the United States. The results also show that Scaled Agile is being applied across different fields, but is dominated by computer science, engineering, and business. We visualized the high-frequency terms using a word cloud and the keyword co-occurrence map, and a density map using VOSViewer. The h-index of 21 for the analysed articles indicates the significant scholarly impact of the publications. The study identified the following key themes: team dynamics, organisational structures, and practical applications of Scaled Agile. The study also identifies the major challenges associated with Scaled Agile, namely cultural issues and scalability issues, effective organisational design, and change management strategies. The findings of this study offer valuable insights into the current state of Scaled Agile that appeal to industry practitioners and academics interested in Scaled Agile research and implementation.

Since anthropogenic causes accelerate rapid climate change with intensifying the adverse impacts of climate induce hazards, Non-State Actors (NSAs) have emerged as pivotal actors in climate governance. The aim of this research is to explore the diverse roles and contributions of NSAs in climate governance and analyze the challenges and institutional barriers they encounter with proposing some recommendations to strengthen their impact. It employs a qualitative approach where data were collected through KII method. Thematic analysis reveals some meaningful role of NSAs in climate governance including advocating for climate justice, raising awareness, promoting sustainable technologies, enhancing community adaptation and resilience, and collaborating across sectors. Digital awareness campaign of Greenpeace during the Copenhagen and Paris Conference and BRAC's climate-resilient housing and rainwater harvesting initiatives in Bangladesh can be placed as notable examples of NSAs’ roles. Despite their significant contributions, several persistent challenges such as poor coordination among NSAs and with state actors, legitimacy deficits, governance gaps, lack of institutional support and insufficient financing impedes them to realize their full potential. To overcome these challenges, this study recommends the need for legal inclusion of NSAs’ roles, inclusive participation, incorporating intersectionality, stronger accountability mechanisms and sustainable financial frameworks. Furthermore, this study offers actionable recommendations for policymakers and practitioners seeking to enhance the effectiveness of non-state engagement in climate action.